Framework for inferring empirical causal graphs from binary data to support multidimensional poverty analysis.

Area of Science:

• Social Sciences
• Statistics
• Public Health

Background:

• Poverty is a complex global issue requiring accurate measurement.
• The Multidimensional Poverty Index (MPI) uses binary indicators from surveys to assess poverty.
• Existing methods lack a framework to understand causal relationships among these binary poverty indicators.

Purpose of the Study:

• To develop a novel framework for inferring causal relations among binary variables in poverty surveys.
• To provide a tool for analyzing empirical causal links between different aspects of poverty.

Main Methods:

• Proposed a new framework to infer causal relationships on binary variables.
• Validated the approach using simulated datasets with known ground truth.
• Applied the framework to real-world datasets, including poverty surveys and the Twin births dataset.

Main Results:

• The proposed framework outperformed baseline methods in simulated data.
• Successfully identified a causal relation in the Twin births dataset.
• Discovered a causal link between smoking and alcohol consumption in a Thailand poverty survey.

Conclusions:

• The framework effectively infers causal relations among binary variables.
• The 'BiCausality' R package offers a versatile tool for causal inference beyond poverty analysis.
• This approach can enhance understanding and intervention strategies for multidimensional poverty.